Research on Evaluation of Scientific and Technological Innovation Efficiency in Chinese Universities-An Analysis Based on Sample Data of 31 Provinces

2011~2020 is a golden decade of innovation-driven development for China. During this period, scientific and technological innovation (STI) of Chinese universities has achieved historic development. To better capture its innovation trajectory and shed light on practical implications, this study uses DEA-BCC model and Malmquist index to calculate the STI efficiency of Chinese universities in 31 provinces and municipalities from 2011 to 2020. The findings illustrate that: 1) the barrier effect of pure technical efficiency is greater than that of scale efficiency; 2) although the STI efficiency of Chinese universities increased by 0.6% on average, the contribution rate of technological progress growth was high, and the scale efficiency showed negative growth; 3) in terms of regional distribution, DEA efficient provinces and cities are concentrated in northern China, while non DEA efficient provinces and cities are mostly concentrated in southern China. Based on the findings, countermeasures and suggestions are made to improve the STI efficiency for Chinese universities from the aspects of scientific innovation alliance, evaluation mechanism and innovation management ability. It provides a theoretical and practical reference for innovation-driven development of Chinese universities.


I. INTRODUCTION
Since the beginning of the 21st century, innovation has become the focus of global competition. Science and technology(S&T) driven innovation from the perspective of regional differences and policy implications in varied fields have become the focus of academic attention. By entering the new century, China, as the world's second largest economy, has made considerable progress in the S&T innovation-driven The associate editor coordinating the review of this manuscript and approving it for publication was Justin Zhang . development, despite of its large regional differences in economic growth, natural environment and demographic features. Although the regional gap in S&T innovation efficiency is found to be the deep reason of China's regional imbalance, the innovation-driven development trend has dramatically boosted the innovation efficiency of Chinese universities, facilitating to China's socio-economic sustainable development. In May 2022, Tongji University, together with Elsevier, released the report ''Accessing Universities' Contributions to Sustainable Development Goals'', which emphasized that Chinese universities are becoming the key force to achieve VOLUME 11, 2023 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ the goal of sustainable development [1]. From 2011 to 2020, the economic development represented by mobile Internet applications entered a new stage driven by innovation, during which the Scientific and Technological Innovation (STI) of Chinese universities achieved historic and leapfrog development, and new progress were made in STI fields [2]. In particular, in the past decade, the number of innovative talents in Chinese universities has continued to grow, the STI capabilities of colleges and universities have been enhanced, and the innovation platform system has been gradually improved, making irreplaceable and key contributions to China's construction of an innovative and strong country.
With the fierce competition of innovation on a global scale, STI of universities plays an increasingly critical role, which attracts more and more research attention on the related topics. Among these hot topics, STI efficiency of Chinese universities is significant to technological innovation and scientific breakthrough, which considerably benefits to the sustainable development in the applied research [3]. With the development of new technology, Chinese universities undertake the social responsibility of knowledge inheritance, innovation and dissemination in the past decade, and has become an essential part of the national innovation system [4], [5].
In the aspect of STI efficiency, a large number of scholars have conducted investigations on the STI efficiency of Chinese universities in different types and from different regions. Firstly, research is implemented on the STI efficiency in different types of universities in China. For example, Zheng et al. [6] used the Malmquist model to explore the STI efficiency trend in six major categories of universities in China from 2009 to 2019, and found that among the six categories of universities in China, only the total factor productivity of normal colleges was greater than 1, and the other five types of universities were less than 1, and the overall STI efficiency of universities showed a downward trend. Ma et al. [7] used the Malmquist model to analyze the efficiency of STI in science and humanities universities, finding that the dynamic change of efficiency of arts and science universities was not significant. Ke and Yao [8] based on CCR model and BCC model, analyzing the STI efficiency of Chinese top universities from 2008 to 2017, the study pointed out that the comprehensive efficiency of the studied universities was 0.6014, and the main constraint was scale efficiency.
Secondly, research is carried out on the STI efficiency of Chinese universities in different regions. For example, Xu [9] and Wu [10] have calculated the STI efficiency of those universities located in the Yangtze River Delta region. The results showed that low scale efficiency is found to be the major reason for the failure of comprehensive efficiency to reach effectiveness, and the decline of production technology is the major reason for the decline in the average annual growth rate of STI efficiency of universities in this region. Zhu et al. [11] used DEA method to study the STI efficiency of universities in various provinces and cities, and found that the efficiency of achievement transformation was higher than that of knowledge innovation, and the improvement of the overall efficiency of STI in Chinese universities mainly depended on the efficiency of achievement transformation; Wang and Chen [12] constructed a two-stage DEA model, taking 31 provinces and cities as research samples, and the results showed that compared with the transformation stage of S&T achievements, there were fewer provinces and cities where DEA was effective in the previous stage. In addition to taking regions as research objects, some scholars also studied the STI efficiency of universities in a single province. For example, Ruan [13] and Song [14] have studied the STI efficiency of universities in Shandong Province and Hubei Province respectively. In these studies, scholars have constructed different index systems to measure the STI efficiency of universities in different types and regions. However, many studies found that the innovation efficiency of Chinese universities is generally at low level especially in the aspect of research and application. Meanwhile, the positive effect of R&D in universities on high-quality economic development is still insignificant [15]. Therefore, studying spatio-temporal evolution of the STI efficiency of Chinese universities and further exploring ways to improve their STI efficiency are very essential to promote innovation-driven and sustainable development of China.
To sum up, at present, scholars have made many achievements in the empirical study on the STI efficiency of universities, but the academic work on the STI efficiency of Chinese universities is still insufficient. First, the output types of innovation achievements are not clearly divided, and the evaluation indicators of STI efficiency in universities are generally constructed from the perspective of overall output of achievements. Second, most studies focus on the static or dynamic STI efficiency of universities unilaterally, without comprehensive research on static and dynamic STI efficiency. In addition, due to the timeliness of the data, there is still a lack of detailed analysis of the distribution characteristics of different efficiency areas in the period of 2011∼2020, and the spatial analysis method still needs to be explored to illustrate the spatio-temporal evolution. Therefore, this paper aims to close the research gaps and capture the innovation trajectory of Chinese universities in the mentioned decade by exploring their spatio-temporal evolution in the STI efficiency. Technically, it uses the DEA-Malmquist model and Geographic Information System (GIS) software ArcGIS to analyze the STI efficiency of universities both from static and dynamic angle in 31 provinces in China during 2011∼2020, which can greatly make up for the deficiencies in the research field.
In structure, this paper first applies the DEA method to analyze the static efficiency of STI efficiency of universities with the BCC model, and explores the main factors affecting the effectiveness of DEA in STI efficiency. Then, with application of the DEA-Malmquist model, the dynamic efficiency research is carried out, and the STI efficiency of Chinese universities is comprehensively evaluated, indicating the changes of STI efficiency in different years and regions. Based on the analysis, countermeasures and suggestions are put forward to promote the STI efficiency of Chinese universities for practical implications.

II. RESEARCH DESIGN A. DATA SOURCES AND INDICATOR SELECTION
When Following the principles of authenticity, completeness and availability of data, 31 Provinces, Autonomous Regions and Municipalities directly under the Central Government of China (PAMCGs) as the decision making unit(DMU), and draws data source from the ''Compilation of S&T Statistics of Colleges and Universities''(CSTS). This Compilation is launched yearly by the S&T Department in the Education Ministry of China according to the overall situation of S&T activities of more than 1,900 colleges and universities in China. The data covers basic and applied research, R&D investment, S&T output, international S&T exchanges, and other S&T services. Therefore, the data source is authoritative and representative. By referring to the evaluation methods of Zheng [6], Ruan [13] and other scholars on the STI efficiency, this paper constructs an comprehensive evaluation index system measuring the STI efficiency of Chinese universities in the PAMs at provincial level (listed in Table 1).
In this system, the STI efficiency of Chinese universities is evaluated from two categories, namely input index and output index. In terms of input index, it consists of three primary indexes, including investment, manpower input and construction of scientific research environment. Among them, the ''financial investment'' may cover all investment in the S&T activities from the finance perspective. To elaborate the investment channel, we use ''R&D expenditure(X1)'' as its proxy. ''Manpower input'' is mainly related to human resources input in the S&T activities, measured by ''staff number in teaching and research(X2)''. As in Chinese universities, most staff have both research and teaching obligations. Hence, we consider the total number of teaching and research staff as the proxy of the S&T manpower input. The ''construction of scientific research environment'' mainly refers to the programs and atmosphere to support S&T activities. For example, most Chinese universities set up varied research fundings for research initiatives. The ''construction of scientific research environment'' is measured by the number of S&T projects(X3) and international exchange(X4) as proxy. As more S&T projects can motivate and involve more staff in S&T innovation activities to build a positive research atmosphere, its quantity can be a proxy of construction of scientific research environment. The ''international exchange'' is indicated by the sum of international cooperation dispatches and international conference hosted.
In terms of output index, it consists of primary index including ''academic achievements'' and ''S&T achievements''. The ''academic achievements'' are measured by ''published S&T works(Y1)'' and ''published academic papers(Y2)''. The indexes can directly present the innovation outcome in the academic fields, which have been chosen as widely applied proxy in the S&T efficiency research. The ''S&T achievements'' are measured by the number of ''patent authorization'' and ''technology transfer contracts''. Although the pure number of patents and technology transfer contracts may not directly indicate the eventual economic or related impact, it can reflect the innovation ability and potential of universities at provincial level, which can act as proper proxy for S&T achievement. Based on the evaluation index, the data corresponding to each indicator comes from the statistical yearbook CSTS from 2011-2020. Furthermore, the processed data are adjusted in format to fit the DEAP2.1 tool for analysis.

B. STATIC EFFICIENCY ANALYSIS BASED ON DEA-BCC MODEL
Data envelopment analysis (DEA) was firstly introduced in 1978 by Charnes et al. [16]. In the field of STI efficiency research, DEA is most widely used [17]. According to the concept of relative efficiency, DEA is considered as a nonparametric statistical method used to evaluate the productivity of decision-making units with the same type of multiple inputs and outputs [18], [19].
The CCR model and BCC model are included in the traditional DEA method. The former one presumes that the return to scale remains unchanged, while the later model supposes that the return to scale is variable, in which case technical efficiency will be affected by efficiency of scale [20], [21].
The mathematical model can be displayed as below: The variable x is the input element, and x ij represents the input amount of the ith input element in the j th decision unit; The variable y is the output factor, and y rj represents the output of the r output factor in the j th decision unit; λj is the coefficient of each unit combination;ε is a non-Archimedes infinitesimal; θ is the efficiency evaluation coefficient, while s-and s+ are the relaxation variables.

C. DYNAMIC EFFICIENCY ANALYSIS BASED ON DEA-MALMQUIST MODEL
The Malmquist index is a productivity measurement method [22]. The traditional DEA model (CCR and BCC) usually explores the production efficiency of the decision-making unit at one time point, but difficult to reflect the dynamic change. The Malmquist exponential model can solve this problem very well, as it is an index that reflects changes in relative efficiency, expressed as the rate of change in total factor productivity (TFPCH), calculated as: The change in the total factor generation rate does not come from the influence of a single indicator, but can also be decomposed layer by layer, which is calculated as follows: Among them, the rate of change of technological progress (TECHCH), the rate of change of technological efficiency (EFFCH), the rate of change of pure technical efficiency (PECH), and the rate of change of scale efficiency (SECH) are calculated as follows:

A. STATIC STI EFFICIENCY ANALYSIS OF CHINESE UNIVERSITIES
The multi-stage DEA with variable output-oriented scale remuneration was used to calculate the STI efficiency of universities in each province in 2011 and 2020, and the analysis results (listed in Table 2) were completed by DEAP2.1 software tools. Shown in the Table 2, when the overall efficiency equals to 1, the decision unit DEA is effective. On the contrary, when the overall efficiency does not equal to 1, the decision unit is not DEA valid; The comprehensive efficiency(crste) can be decomposed into pure technical efficiency(vrste) and scale efficiency(scale), when the technical efficiency or scale efficiency equal to 1, it means that the efficiency is effective, otherwise the efficiency is invalid.
Meanwhile, the type of scale remuneration of the decisionmaking unit can also be judged according to whether the scale efficiency equals to 1: if the scale efficiency is equal to 1, the decision-making unit is in the stage of unchanged scale remuneration; otherwise the decision-making unit is in the stage of increasing or decreasing return to scale [23]. Based on the calculations of the equation (1)∼(6), the average comprehensive efficiency of universities in China in 2011 and 2020 were 0.919 and 0.921, the average technical efficiency was 0.924 and 0.950, and the average scale efficiency was 0.994 and 0.969, respectively, indicating that the overall innovation efficiency of Chinese universities needs to be improved. In 2011, there were 17 PAMCGs with effective DEA, accounting for 54.8%, 18 PAMCGs with effective technical efficiency, accounting for 58.1%, and 17 PAMCGs with effective scale efficiency, accounting for 54.8%; In 2020, there were 13 PAMCGs with effective DEA, accounting for 41.9%, 17 PAMCGs with effective technical efficiency, accounting for 54.8%, and 14 PAMCGs with effective scale efficiency, accounting for 45.2%. In summary, whether in 2011 or 2020, there are more PAMCGs with effective technical efficiency than those with effective scale efficiency, but the technical efficiency is smaller than the scale efficiency, indicating that the technical efficiency of universities in various PAMCGs has decreased significantly, and the hindrance effect of PECH is more significant. At the same time, compared with 2011, the PAMCGs with effective DEA, effective technical efficiency and effective scale efficiency decreased by 4, 1 and 3 respectively in 2020, specifically: 1) In the year of 2011 and 2020, there were 10 PAMCGs with effective DEA, including Hebei, Inner Mongolia and Jiangsu, accounting for 32.3%. The technical efficiency and scale efficiency of these provincial and municipal universities are effective in 2011 and 2020, the rational allocation and utilization of resources, the optimal scale of innovation, and there is no input redundancy or insufficient output. 2) In 2011, non-DEA was effective, and in 2020, there were 3 PAMCGs with effective DEA, Shanxi, Chongqing and Guizhou, accounting for 9.7%. In 2011, the technical efficiency and scale efficiency of Shanxi and Guizhou were less than 1, and the technical efficiency was less than the SECH, while the technical efficiency of Chongqing was equal to 1, and its non-DEA effectiveness was caused by the ineffective scale efficiency. 3) In 2011, VOLUME 11, 2023 the DEA was effective, and in 2020, there were 7 PAMCGs that were not DEA effective, including Liaoning, Shanghai, and Anhui, accounting for 22.6%. In 2020, only Tibet has a technical efficiency equal to 1, and the scale efficiency is invalid, resulting in its non-DEA effectiveness, and the technical efficiency and scale efficiency of other PAMCGs are less than 1, and the technical efficiency is less than the scale efficiency. 4) In 2011 and 2020, there were 11 PAMCGs that were not DEA valid, including Beijing, Tianjin and Jilin, accounting for 35.5%. In 2011, the technical efficiency and scale efficiency of these PAMCGs were less than 1, and the technical efficiency was less than the scale efficiency; In 2020, the technical efficiency of Heilongjiang, Hunan and Guangdong is equal to 1, and the reason why the non-DEA is effective is that the scale efficiency is invalid, and the technical efficiency and scale efficiency of other PAMCGs are less than 1, and the technical efficiency is less than the scale efficiency.
From the perspective of scale remuneration types, in 2011, Chinese universities were in 17 PAMCGs with unchanged scale remuneration, accounting for 54.8%, 8 PAMCGs with increasing scale remuneration, accounting for 25.8%, and 6 PAMCGs with decreasing scale remuneration, accounting for 19.4%; In 2020, Chinese universities were in 15 PAMCGs with unchanged scale returns, accounting for 48.4%, 3 PAM-CGs with increasing scale returns, accounting for 9.7%, and 13 PAMCGs with decreasing scale returns, accounting for 41.9%. In 2020, the PAMCGs with unchanged scale returns and increasing scale returns decreased by 6.4% and 16.1% respectively compared with 2011, and the PAMCGs with decreasing scale returns increased by 22.5%, specifically: In 2011, there were 8 PAMCGs with increasing scale returns, including Tianjin, Shanxi and Jilin, and only Tianjin, Tibet and Ningxia increased scale returns in 2020, of which Tianjin was in increasing scale returns in 2011 and 2020, indicating that the S&T activities of universities in these PAMCGs were insufficient. It is in need to improve the investment in STI resources and talents, strengthen the construction of STI environment, and improve scale efficiency to achieve the optimal scale. In 2011, there were 6 PAMCGs with decreasing scale returns, including Beijing, Jiangxi and Shandong, and 13 PAMCGs with decreasing scale returns in 2020, including Beijing, Liaoning, and Heilongjiang, of which 6 PAMCGs with decreasing scale returns in 2011 were also in a state of decreasing scale returns in 2020, indicating that universities in these PAMCGs have the problem of excessive scale investment, and should reasonably reduce the scale or optimize and adjust the scale.

B. THE CHANGING TREND OF STI EFFICIENCY OF STI EFFICIENCY OF CHINESE UNIVERSITIES
As demonstrated in Table 3, the output-oriented Malmquist index model was applied by DEAP2.1 software tools to calculate the comprehensive STI efficiency of universities in various PAMCGs during 2011∼2020. Because the Malmquist model reflects the relative efficiency change from decision unit t to t+1, and the data of 2011 as the initial year cannot be compared, the table only contains the efficiency change from 2012 to 2020.
According to Table 3, the average value of the Malmquist index (hereinafter referred to as the M index) of China's universities in 2011∼2020 is 1.006, with an average increase of 0.6%, indicating that the efficiency of STI in Chinese universities is growing and shows a relatively stable level of STI development. Since 2012, STI in Chinese universities has ushered in a new stage of development, and the Chinese government has made major progress for the implementation of innovation-driven development strategies, and put STI at the core of the overall national development [24]. In order to create an innovative atmosphere and increase the confidence of innovation subjects, China has released a large number of influential policies, focusing on strengthening discipline construction, cultivating innovative talents, improving the reform of S&T evaluation, clarifying the important position of universities in the national STI system.
In terms of M-index decomposition, the average value of technological progress (TECHCH) is 1.006, with an average increase of 0.6%, compared with EFFCH, the growth contribution rate of technological progress is higher, indicating that the growth of STI efficiency of Chinese universities mainly depends on technological innovation and progress. The average value of EFFCH is 1.000, the efficiency has not changed, by decomposing the index, the PECH is 1.003, with an average increase of 0.3%, indicating that the allocation and utilization efficiency of university innovation resources has been improved, but its role in promoting is not significant and still needs to be improved. The EFFCH has not changed, mainly caused by the decline in SECH, the average SECH is 0.997, with an average decrease of 0.3%, the overall downward trend, is the main factor hindering the improvement of EFFCH. However, from the perspective of each year, the fluctuations in SECH are not large, but basically keep stable, indicating that the change in the scale of STI in Chinese universities has no significant effect on the improvement of efficiency, and there is still much room for improvement. Compared with the technology progress index, the change trajectory of technological progress and M index is very similar. Compared with other efficiency indexes, technological progress has changed a lot. In 2013, it decreased by 0.199 compared with 2012, and later increased by 0.187 until 2016. The instability of the technological progress index has affected the stability and improvement of the STI efficiency of Chinese universities to a certain extent. However, the EFFCH is affected by the PECH index and the SECH index, and the change range is small, with the maximum value at 1.035, and the minimum value at 0.980. The increase and decrease change are relatively stable, and its impact on the change of the M index is limited.
By comparing the overall development trend of Effch, Techch, Pech, Sech and Tfpch shown by M index from 2012 to 2020 (depicted in Figure 1), it can be seen that Techch has the largest change range, followed by Tfpch. The changes of Effch, Pech and Sech are relatively stable. It demonstrates that in the year of observation, the technical progress and total factor productivity have changed greatly, but the fluctuation range of EFFCH, PECH and SECH are relatively small.

C. ANALYSIS OF STI EFFICIENCY OF UNIVERSITIES IN VARIOUS PAMCGS
The empirical results illustrate ( Table 4) that in 2011∼2020, the M-index of STI efficiency of universities in 18 PAMCGs obtained positive growth, of which the efficiency of universities in 10 PAMCGs increased by more than 5%, among which the STI efficiency of Guizhou (1.100), Zhejiang (1.091) and Heilongjiang (1.061) increased by 10.0%, 9.1% and 6.1% respectively, showing a high level of STI. Guizhou is mainly affected by technological progress (1.070), and its technological progress index ranks second in the country. In addition, Guizhou's EFFCH (1.027) is also far higher than the national average (1.000), indicating that Guizhou's universities have significant technological progress and innovation efficiency, improved resource utilization efficiency, and formed a certain scale benefit. Zhejiang's technological progress (1.091) increased most significantly, but the EFFCH (1.000) did not improve. Heilongjiang's technological progress (1.041) and EFFCH (1.019) both improved, but the SECH (0.999) decreased, hindering the improvement of its overall efficiency. In addition, there are 13 PAMCGs with an M index less than 1, among which Tibet (0.883), Ningxia (0.893) and Xinjiang (0.922) have the most significant declines, mainly caused by a sharp decline in technological progress, and at the same time, EFFCH has not been well improved.
In terms of technological progress index, the technological progress of universities in 18 PAMCGs is greater than 1, among which Zhejiang (1.091), Guizhou (1.070) and Jiangxi (1.062) have the largest increase in technological progress, reaching 9.1%, 7.0% and 6.2% respectively. The technological progress of universities in 13 PAMCGs is less than or equal to 1, and the technological progress of universities in Ningxia (0.917), Tibet (0.919) and Xinjiang (0.922) has decreased the most, with decrease rate at -8.3%, -8.1% and -7.8% respectively.
In terms of EFFCH index, the EFFCH of universities in 12 PAMCGs is greater than 1, which is consistent with the number of PAMCGs with PECH greater than 1. Among the 12 PAMCGs, the PECH is greater than the SECH in 11 PAMCGs, and only the SECH in Chongqing (1.002) is higher than the PECH (1.000). PAMCGs with high EFFCH, such as Tianjin (1.045), Guangdong (1.040), Jilin (1.032), etc., also rank high in PECH; PAMCGs with low EFFCH, such as Fujian (0.944), Anhui (0.966), Ningxia (0.974), etc., are also not ideal in PECH. The main reason is that the change in SECH is small, and there is a downward trend in most PAMCGs, which is difficult to have a greater impact on EFFCH. Only 3 PAMCGs in the country have a SECH greater than 1, namely Chongqing, Guizhou and Shanxi, with an increase rate at 0.2%, 0.1% and 0.1% respectively. And among the remaining 28 PAMCGs, the SECH of 14 PAMCGs is equal to 1, and the SECH of 14 PAMCGs is less than 1, and the overall SECH is relatively low. However, the positive and negative changes in the SECH of universities in 31 PAMCGs are small, only Tibet (0.961) and Shandong (0.988) have a change in SECH of more than 1%, which shows that Chinese universities have scale-expansion while failing to reasonably optimize the scale structure, so that the efficiency improvement brought about by the scale change is small or negative.

D. REGIONAL CHARACTERISTICS OF STI IN CHINESE UNIVERSITIES 1) REGIONAL CHARACTERISTICS OF STATIC STI EFFICIENCY
Based on the static efficiency results of STE of Chinese universities in various PAMCGs in 2011 and 2020, ArcGIS10.4 is used to draw a spatial visualization map (depicted in Figure 2), which better presents the regional aggregation and changes over time of effective and noneffective DEA of Chinese universities in various provinces and cities in 2011 and 2020 respectively.
On the whole, obvious regional differences in the static STI efficiency of Chinese universities are displayed in Figure 2. The PAMCGs with effective DEA (the comprehensive efficiency is equal to 1) are roughly clustered in the northern region of China (north of the Qinling Mountains and Huaihe River, including the northern region, northwest region, and Qinghai-Tibet region). The PAMCGs with effective DEA in 2011, such as Inner Mongolia, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang, as well as the PAMCGs

2) REGIONAL CHARACTERISTICS OF DYNAMIC STI EFFICIENCY
Based on the dynamic STI efficiency results of Chinese universities in various PAMCGs from 2011 to 2020, we take the efficiency value of 1.000 as the basis for the classification of high and low efficiency. To be specific, if the value of EFFCH or TECHCH is greater than 1, the quality of STI development is considered in the high efficiency stage. On the contrary, if the value of EFFCH or TECHCH index is less than or equal to 1, it is considered that the development of STI is slow or stagnant, and it is in a low efficiency stage. According to the above rules, a scatter map of ''effch-techch'' of Chinese universities is drawn. It divided 31 PAMCGs into four quadrants, namely ''high-high'' efficiency, ''low-high'' efficiency, ''low-low'' efficiency and ''high-low'' efficiency. ArcGIS10.4 is used to draw a spatial visual map to explore the distribution characteristics of the STI efficiency of Chinese universities in various PAMCGs (depicted in Figure 3).
As shown in Figure 3 and Figure 4: The first quadrant is distributed with PAMCGs with ''high-high'' efficiency, that is, high EFFCH and high technical progress. The development of STI in these PAMCGs is relatively balanced. While promoting scientific and technological development, resource allocation and scientific research management capabilities are also improving, effectively promoting the  The second quadrant is distributed by PAMCGs with ''low-high'' efficiency, that is, low EFFCH and high technical progress. These PAMCGs are dominated by technical progress, with inadequate resource interconnection management, low resource allocation efficiency and inadequate financial control. As shown in Figure 3 The third quadrant is distributed by PAMCGs with ''lowlow'' efficiency, that is, low EFFCH and low technical progress. Both EFFCH and technological progress are inefficient, which seriously hinders the development of STI. These PAMCGs are widely distributed, without significant aggregation. Shanghai, Fujian and Hainan are located in the east, while Henan and Hubei in the middle region, Tibet, Qinghai, Ningxia and Xinjiang in the west, and Liaoning in the northeast.
In the fourth quadrant, the PAMCGs with ''high-low'' efficiency are distributed. These PAMCGs are dominated by EFFCH, and technological progress is declining. Only Beijing, Shanxi and Jilin belong to this type, and their spatial distribution is relatively dispersed.

A. RESEARCH FINDINGS
The DEA-Malmquist index model is applied in this study to empirically analyze the STI efficiency of universities in China during 2011∼2020, combined with the GIS analysis method via ArcGIS software to analyze the changes in STI efficiency spatially. Three main findings are summarized in the following session: The results of static analysis illustrate that the overall lack of STI efficiency of Chinese universities in 2011 and 2020 is more significant than that of scale efficiency. In 2020, China's PAMCGs with effective DEA, effective technical efficiency and effective scale efficiency decreased compared with 2011, and in terms of comprehensive efficiency: 10 PAMCGs with effective DEA in 2011 and 2020, accounting for 32.3%; 3 PAMCGs with non-DEA effective in 2011 and DEA effective in 2020, accounting for 9.7%; There are 7 PAMCGs with effective DEA in 2011 and non-DEA in 2020, accounting for 22.6%; In 2011 and 2020, there were 11 PAMCGs that were not DEA effective, accounting for 35.5%. At the same time, in the PAMCGs where the DEA is not effective, the value of technical efficiency of most studied provinces is less than that of scale efficiency. From the perspective of types of scale compensation: 8 provinces and 3 cities with increasing scale returns in 2011 and 2020 respectively, of which Tianjin increased in both 2011 and 2020; In 2011 and 2020, there were 6 provinces and 13 PAMCGs with diminishing returns to scale, respectively, of which the PAMCGs with diminishing returns on scale in 2011 were also in a state of diminishing returns on scale in 2020.
The dynamic analysis results show that the average growth of the M index of STI efficiency of Chinese universities in 2011∼2020 is 0.6%, demonstrating a positive growth trend, the growth of the M index mainly depends on technological progress, and the change trajectory of the two is very similar, while the growth contribution rate of EFFCH is low, and the impact on the M index is small, among which the SECH is negative growth, which hinders the improvement of STI efficiency. In terms of regions, the M-index of STI efficiency of 18 PAMCGs has a positive growth, and a total VOLUME 11, 2023 of 13 provincial and municipal universities with negative growth of M-index; Technological progress increased by an average of 0.6%, consistent with the number of PAMCGs with positive or negative growth on the M index; In terms of EFFCH, 12 PAMCGs showed positive growth, 8 PAMCGs showed negative growth, and 10 PAMCGs showed efficiency equal to 1, mainly because the PAMCGs with positive growth in SECH were only Chongqing, Guizhou and Shanxi, and the other 28 PAMCGs had SECH less than or equal to 1, and the overall SECH was low.
Based on the static efficiency results, the PAMCGs with effective DEA in China are roughly distributed in northern China, while the PAMCGs with non-DEA effectiveness are mostly distributed in the southern region. Based on the dynamic efficiency results, Based on the dynamic efficiency results, the PAMCGs with ''high-high'' efficiency are mainly distributed in the central and western regions, showing the characteristics of large concentration and small dispersion; Most of China's ''low-high'' efficiency PAMCGs are concentrated in the central and eastern and western regions, showing a multicenter distribution characteristics; The PAMCGs with ''low-low'' efficiency are widely distributed, and the clustering feature is not significant, showing a uniform distribution feature; The PAMCGs with ''high-low'' efficiency are only Beijing, Shanxi and Jilin, which are distributed dispersedly.

B. LITERATURE CONTRIBUTIONS
This study has made up for the deficiencies in the research field and has the following three main contributions. Firstly, it has built a comprehensive evaluation system of the STI efficiency of Chinese universities at provincial level, combining both input and output categories for their dynamic performance. When evaluating the STI efficiency of universities at provincial level, most previous studies focused more on the overall STI efficiency on the static frontier performance. This study has more carefully broken down the important parts of the STI. Instead of overall input and output of STI, we elaborate the input category into R&D expenditure, staff number in teaching and research, number of S&T projects and international exchange as the proxy of financial investment, manpower input and construction of scientific research environment respectively. The chosen indicators reflect the realistic situation of Chinese universities based on the data availability in the official statistics. The output types of innovation achievements are clearly divided into published works, published academic papers, number of patent authorization, and technology transfer contracts. Although these indicators may not directly reflect the eventual economic or related impact, they can to a large extent represent the potential innovation-driven development ability of universities at provincial level and act as nice proxy of innovation output. Furthermore, by combining the DEA model and the Malmquist index, a dynamic performance of these proxy indexes is presented, which is helpful to propose specific measures from the perspective of STI efficiency of universities, and to help implement and adjust the strategy of rejuvenating the country through science and education timely.
Secondly, this study combined the static and dynamic STI efficiency of universities for comprehensive evaluation. Innovation is the engine for high-quality economic and social development in a dynamic evolution process. In recent years, the output of STI in China has been expanded, but it is still at relatively low level, that low quality of innovation takes place frequently [25]. The presented study measured the STI efficiency of Chinese universities and gave analysis of spatial-temporal characteristics during 2011∼2020. The findings provide a novel reference by both static and dynamic evolution process of provincial STI efficiency that help guide provinces to take updated innovation activities to promote innovation efficiency and quality.
Thirdly, GIS analysis with software ArcGIS is used in this study to explore the spatial-temporal characteristics of STI efficiency of Chinese universities in different provinces, expanding the scope of innovation efficiency research. Lu et al. argued that China's STI efficiency shows a strong positive spatial correlation [26]. This paper studies the STI efficiency of Chinese universities from the perspective of spatial visualization, which is conducive to further exploring the determinants affecting the STI efficiency of Chinese universities, and provides a reference for relevant departments of national and local governments to formulate science and technology policies.
Compared to the previous research, the results of this study are consistent to those scholars such as Xu et al. [9], Wu and Cui [10], and Song and Zou [14], finding that the main factor restricting the STI efficiency of Chinese universities is low scale efficiency. For example, Xu et al. investigated the STI efficiency of 82 universities in the Yangtze River Delta region during 2013∼2017 [9], Wu and Cui conducted another similar research based on the data of 41 cities in the Yangtze River Delta region during the same period [10], they all found that low scale efficiency is the main reason why the comprehensive STI efficiency of Chinese universities is not effective. Song and Zou carried out another research on Hubei universities' technology innovation efficiency based on 2011∼2013 data, pointing out that the diminishing returns to scale led to low STI efficiency [14]. Although these similar studies have been conducted in different region and across different time period, they have obtained the similar finding results, indicating the credibility and continuity of this presented study. Nevertheless, this study keeps pace with the times and covers a wider range of space to extend the literature.
Compared to studies focusing on spatial distribution of STI efficiency of Chinese universities [7], [8], [13], the results of this study support the fact that the regional differences in innovation efficiency of Chinese colleges and universities are huge. However, most of the previous studies only focused on a certain type of university, without data covering the whole country. For example, Ke and Yao studied the STI efficiency of 140 ''Double First-Class'' universities in 31 provinces of China during 2008∼2017, finding that the innovation efficiency of different regions in China has agglomerated characteristics [8]. Ma et al. compared the STI efficiency of Chinese universities in different discipline distribution in 31 provinces during 2009-2018, illustrating that compared with science and technology universities, the innovation output of humanities and social sciences universities is insufficient [7].
Different from previous studies showing the innovation efficiency of Chinese universities decreased from west to east, and from east to central regions [11], the dynamic efficiency results of this study find that the ''high-high'' quadrant includes increasing provinces in both central and western regions during 2011∼2020. The reason for this difference may be that under the national strategy of ''Western Region Development'' and ''Rise of the Central Region'', the S&T innovation support policies and the innovation inputs in China's central and western regions continue to increase, making the STI efficiency of colleges and universities in these regions continue to improve. It indicates that China's innovation-driven development strategy has made great achievements in the past decade, and the differences between regional development are continuously shrinking.
While verifying the conclusions of the previous research, this study further combines the DEA-Malmquist model and GIS analysis method, and deeply explores the temporal and spatial differences of STI efficiency of Chinese universities through the spatial analysis in 31 provinces in China from 2011 to 2020. Combining the dimensions of both time and space, this study has greatly enriched the literature perspective of the research field in university STI efficiency.

C. PRACTICAL IMPLICATIONS
According to the findings, this paper puts forward three major practical implications. Firstly, it is necessary to establish a STI alliance for research cooperation and application that integrates government, industry, universities and social institutions. Research cooperation is essential to accelerate knowledge innovation and transformation, which has become an important driving factor to promote STI efficiency of universities and enhance provincial competition. The innovation alliance is the main component of strategic scientific and technological forces. It should give full play to the promoting role of the government as leader, the application role of the enterprises as the main body of industrialization and marketization, and the research role of the universities as the engine of knowledge creation and transferring. With the close participation of government departments, it is a task oriented and systematic way to effectively organize the upstream and downstream advantageous enterprises, scientific research institutions and universities in the industrial chain to tackle key problems. To solidate the alliance, it needs to construct resource foundation of sharing service of platform to boost the data exchange and sharing, and accelerate the management of resource interconnection [27]. Meanwhile, on the regional level, cooperation mechanism can be established to promote the innovation factors flow between regions, and strengthen the connection between government, industry and universities. Based on the STI alliance, further measures should be jointly carried out to promote the transformation and application of R&D achievements from universities.
Secondly, the performance evaluation mechanism of STI in colleges and universities should be changed from focusing on quantity and scale to focusing on quality, guiding universities to establish a more scientific research valuation system. Since universities act as subject of knowledge creating and output end in the process of innovation [28], it is necessary to establish a more scientific evaluation system, facilitating to improve the quality of scientific research achievements and strengthen the ability to transform STI as the core. A comprehensive evaluation method should be conducted combining process and results, improving scientific research policy support, and enhancing the ability of STI transformation by motivating scientific research talents. At present, Chinese universities are at the stage of scarcity in human capital and abundance in material capital [29]. Chinese universities should focus on accumulating human capital by optimizing evaluation index. Aiming at the major national strategies and the frontier development direction of disciplines, we should vigorously cultivate highly qualified talents in urgent need, formulate interdisciplinary talent training programs, and explore new mechanisms for high-level talent training. In order to increase the proportion of scientific researchers with senior professional titles, it should promote scientific research stations and attract more young and outstanding scientific research talents, and establish a sound, effective and reasonable incentive, assessment and professional title evaluation mechanism. We should implement a dynamic evaluation and incentive system of matching contributions, eliminate the performance evaluation method of title, qualification and thesis, and measure the level of human capital in an all-round way. Meanwhile, we should pay attention to the academic and economic value of innovation output, optimize the academic ecology, and stimulate the talents' innovation potential and boost their enthusiasm for STI activities.
Finally, the innovation management ability of Chinese universities needs to be further improved. Luo et al. found that the innovation scale efficiency of universities is relatively low, which cannot adapt to China's economic transformation to high-quality level [15]. In the university administrative work, we should innovate the administrative objectives. Because the competition between colleges and universities in provinces is becoming increasingly fierce. Only by adapting the external new situation and changes, timely and effectively adjusting the innovation strategies, and improving the administrative management ability, can we improve the STI efficiency and competitiveness of universities. In addition, improving the innovation ability of university administrative personnel is the necessary premise and fundamental guarantee. As administrative personnel are the main executors of VOLUME 11, 2023 management work in colleges and universities, their innovation ability directly affects the universities' innovation level. Administrators should keep learning to improve their professional level, stimulate their creativity, and improve the comprehensive quality, so as to raise the administration efficiency and create an innovative, efficient and open campus atmosphere. In this way, it can further promote S&T activities and achieve the best innovation effect for Chinese universities.

V. CONCLUSION
There is still a regional gap in the STI efficiency among Chinese universities, and combined effective model and method are urgently needed to evaluate the university STI efficiency timely at provincial level. This paper uses DEA-BCC model and Malmquist index to explore the STI efficiency of Chinese universities in 31 PAMCGs from 2011 to 2020. The conclusions provide a new viewpoint for literature contributions and practical implications, benefit to combine STI efficiency with regional development, and give a scientific reference for the government and universities to adjust future innovation-driven development policies. However, some limitations inevitably exist that provides avenues to explore further research. Firstly, in the selection of evaluation indicators for STI efficiency in universities, with the emergence of innovative achievements generated by digital technology, the indicators in the evaluation system construction in future research can also be expanded. Secondly, the efficiency of S&T activities dramatically depends on the type of university and its discipline structure. For example, comprehensive universities are normally adept at basic research, while science universities are more efficient in applied research [3]. This paper does not differentiate the types of universities, but only considers the STI efficiency of all universities in a single province. Therefore, future research can further subdivide different types of universities in China, and classify their STI efficiency characteristics to put forward more tailored countermeasures and suggestions.